Analysis of asymmetries in the nexus among clean energy and environmental quality in Pakistan

Abstract

This study examines the short-run and long-run asymmetric effects of clean energy consumption on carbon emission in Pakistan, over the annual time period 1975–2018, by using a non-linear ARDL approach. The findings of the study confirm the existence of asymmetries, in the nexus between the clean energy consumption and carbon emission in the short and long run. The findings of non-linear model confirm that carbon emission responded contrary to positive shocks of energy variables as compared with their negative shocks. Asymmetric findings recommend that positive and negative shocks of the alternative and nuclear energy and combustible and waste energy have affected differently. Although, short- and long-run results suggest an insignificant positive and negative relationship between electric power consumption and carbon emissions. Therefore, more taxation of non-renewable energy and clean energy supports are suggested for the Pakistan economy. We concluded that Pakistan has potential in clean energy which will improve environmental quality in the near future.

Introduction

In Pakistan, urbanization and growth of the population have led to energy demand. This tendency has also been observed in the last few decades in Pakistan. However, in order to get sustainable economic growth, the demand for energy is quite high which has threatened the natural environmental balance (e.g., global warming, deforestation, air pollution, and water pollution) in Pakistan. The growing level of unclean energy consumption in Pakistan has led to initiate an investigation of its impact on carbon emissions that may contribute to global warming because the mode in which we consume energy impacts society’s environmental quality. Moreover, macro-instability and oil shock of 2008 affect the price of fuels and increasing greenhouse gases inspired to find clean energy sources including nuclear, biomass and biofuel, and electric power energy. Therefore, the Pakistan government has been following environmental policies to encourage solar, nuclear, biomass and biofuel, and electric power energy.

The clean energy sources are biomass, biofuels, hydropower, geothermal, wind, nuclear, solar, and sea wave energy, respectively. Therefore, clean energy might decrease energy dependency and energy security, and improve environmental quality. Besides, clean energy offers a solution for natural problems of climate change, global warming, acid rains, loss of biodiversity, air pollution, and water pollution because clean energy might slacken carbon emissions and other pollutant gas emissions (Georgescu et al. 2011; Apergis and Payne 2012; Danish et al. 2017; Shahbaz et al. 2017; Zoundi 2017; Ullah et al. 2020). Thereby, clean energy can increase employment and improve living standards by making the economy better, and thus, it can lessen poverty in developing economies (Danish and Wang 2019). Furthermore, clean energy can enhance sustainable economic growth through improvements in agricultural, industrial, and service growths. It is therefore not surprising to see that these severe concerns over rising fossil fuel prices, macro-instability, energy security, and pollutant emissions have carried the significance of clean energy to the broader issue of the energy and environmental debate.

In past research, numerous studies debated that clean energy is the main engine for environmental quality in distinct countries of the globe by using various econometric tools, for example, Danish et al. (2017) for Pakistan, Katircioglu (2015) for Turkey, Dong et al. (2017) for China, Iwata et al. (2010) for France, Bilgili et al. (2016a, b) and Shahbaz et al. (2017) for the USA, Dogan and Seker (2016) for Europe, Bilgili et al. (2016a, b) for OECD, Sebri and Ben-Salha (2014) and Dogan and Inglesi-Lotz (2017) for BRICS countries, Zeb et al. (2014) for SAARC countries, Zoundi (2017) for African countries, and Solarin et al. (2018) for developing and developed countries. However, there has been a lot of studies to inspect the impacts of clean energy on carbon emissions in the globe. These research studies use only a linear model and mainly ignored the non-linear relationships among the clean energy and carbon dioxide emissions.

In sum, previous studies deliberated the impact of unclean energy consumption on carbon dioxide emissions of the globe. The link between unclean energy consumption and environmental degradation was observed very frequently in Pakistan. But there is limited work on the relation between clean energy consumption and CO₂ in Pakistan. Some empirical studies provide the association between clean energy and CO₂ that gives biased results because they used symmetric ARDL. For that reason, this study aims at fulfilling this gap by examining the asymmetric relationship among the consumption of clean energy and CO₂ in Pakistan by using data from 1975 to 2018 that has not been studied before. The foremost purpose of this study is to inspect the asymmetric impact of clean energy, i.e., nuclear and biomass on CO₂ in Pakistan. This study provides the effect of positive (or negative) shock of clean energy consumption on CO₂ in Pakistan. This study, compared with previous studies, has a parsimonious model. Additionally, this is the leading study on Pakistan and the globe that reflects asymmetric effects that provide a new framework in environmental economics. The results provide necessary economic implications for making the environment friendly to civilians, environmentalists, policymakers, researchers, and government authorities. This research is primarily essential within the framework of Pakistan’s vision 2025 and 2035.

Next, the “Literature review” section gives a complete literature review and also provides a summary of the literature. “Methodology and data” shows the methodology, variable definition, and data descriptive statistics. “Results and discussion” gives the symmetric and asymmetric ARDL results of short and long run with economic implications. “Conclusion and policy implications” concludes the paper with some implications.

Literature review (Table 1)

The body of the literature relevant to the current study can be categorized into three sets. The first group aims to explore the empirical literature on renewable energy consumption and carbon emission nexus. A plethora of studies that explore the dynamic relationship between carbon dioxide emission and renewable energy consumption (REC) are accessible. The pertinent literature can be classified into panel data studies and time-series studies. Furthermore, the strand of panel data studies is twofold: firstly, the studies that deduce that the emission of carbon dioxide decreases on the account of using the REC, and secondly, the studies that support the positive association between carbon pollution. As for the first bunch, the majority of researches result in the negative association between REC and CO₂ emission. For instance, Apergis and Payne (2012), employing the panel ECM, explore that REC plays a significant role in decreasing the level of CO₂ emission in selected 80 economies in both short as well as long run. Similarly, Zeb et al. (2014) apply the FMOLS techniques and confirm that, in SAARC economies, the ratio of CO₂ emission faces the downfall as more renewable energy is consumed. Also, Sebri and Ben-Salha (2014) conclude the same findings in the case of BRICS economies.

Table 1 Summary of literature

Moreover, for using the Central American nation’s studies, Apergis and Payne (2014) infer that, in the long run, REC mitigates the detrimental repercussions of CO₂ significantly after 2002 as compared with pre-2002. Likewise, for the European Union (EU) economies, the findings of the panel DOLS approach by Dogan and Seker (2016) confirm that REC and trade alleviate the harmful effects of toxic gases. The same result is also reported by Jebli et al. (2016) for selected 25 OECD economies, applying panel FMOLS method. Another study by Bilgili et al. (2016a, b) for selected 17 OECD countries employs panel FMOLS and DOLS approaches and supports the findings of Jebli et al. (2016). Besides, Bhattacharya et al. (2017) observed 85 developed and developing economies and deduce that the ratio of carbon dioxide significantly reduces due to consuming clean energy. Bhattacharya et al. (2016) also reported that government policymakers are increasing clean energy investment in 38 countries. Likewise, Paramati et al. (2017) find out the same results for G-20 nations. Moreover, taking the data for 25 African economies, Zoundi (2017) explores the significantly negative nexus between the REC and CO₂ emission. Ahmed and Ahmed (2018) and Ahmed et al. (2019) revealed that REC remained significant in the pollution emission in the short and long run in developing countries. Shahbaz et al. (2015) have also advocated in favor of clean and green technologies in Australia. Ahmed and Ozturk (2018) shows that technical innovation increases clean energy intensity in China. A similar result is also found in the case of Malaysia by Shahbaz et al. (2016).

As far as the second bunch is concerned, Apergis et al. (2010), employing the panel ARDL technique, noted that REC effects the CO₂ in selected 19 developing and developed economies. In addition, taking the data for 5 North African economies, Jebli and Youssef (2017) also confirm that REC causes more CO₂ emission. As for the strand of time-series data studies, again, these are twofold. First is the studies that proclaim the negative nexus between REC and CO₂ emission. Jaforullah and King (2015), for the US economy, infer that, in both periods, REC creates a negative impact on the emission of CO₂. Likewise, Danish et al. (2017) also report the same results in the case of Pakistan. Secondly, two pieces of evidence indicate no significant connection between the production of carbon and REC, viz., Menyah and Wolde-Rufael (2010) for the USA, and Lin and Moubarak (2014) for China. Summing up, ample literature reveals the mix results regarding the effects of REC on CO₂ emission. Furthermore, there are only two studies, i.e., Zeb et al. (2014) for SAARC economies including Pakistan and Danish et al. (2017), which focused on Pakistan’s economy.

While, the second group focuses on investigating nuclear energy consumption and CO₂ discharge nexus. Compared with the first bunch of studies, many researchers explore nuclear energy consumption (NEC) and the emission of CO₂ nexus. The available research can be categorized into panel data research and time-series research. As for the panel data studies, it can be further divided into two bunches. The first bunch contains the studies that support the argument that NEC significantly contributes to purifying the atmosphere by reducing the carbon dioxide emission. For instance, Richmond and Kaufmann (2006) deploy the random effect model and conclude that the trend of CO₂ falls due to NEC in the case of OECD economies. However, in the case of non-OECD economies, NEC exhibits no notable impact on the reduction of carbon dioxide emissions. Similarly, Apergis et al. (2010) demonstrate that, in the case of 19 developing and developed economies, NEC decreases the unfavorable environmental effects of CO₂ emission. Also, Baek (2015) reports that, in 12 primary nuclear energy-–generating economies, the level of carbon dioxide emission declines as more of NE is utilized. While, the second bunch includes the studies that indicate the insignificant impact of NEC on the emission of CO₂, for example, Al-Mulali (2014) for 30 major NE-consuming economies, Baek and Pride (2014) for the top six nuclear energy–generating economies, Saidi and Mbarek (2016) for 9 selected developed nations, and Ozturk (2017) for 9 Latin American countries.

As far as the second group is concerned, employing the time-series data for France, Iwata et al. (2010) suggest that NEC promotes the decrease in the ratio of CO₂ emission. On the contrary, Jaforullah and King (2015) gather that NEC carries no significant dynamic association with carbon emissions in the USA. In short, the available relevant studies report the mix findings regarding the nexus of NEC and CO₂ emission. Besides, apparently, there is not a single study in the context of Pakistan that investigates the effects of NEC on the reduction in carbon dioxide. Further, the third group lists all the studies that report the biomass energy and carbon emission nexus. Researchers also explore the dynamic influence of biomass energy consumption (BEC) on carbon emission, and these studies can also be classified into two groups. The first group of scholars demonstrates that the role of BEC is significant in decreasing environmental stress by mitigating the emission of carbon dioxide. As Katircioglu (2015) applying the bound testing approach concludes, the trend of CO₂ emission faces downfall due to BEC in Turkey. Likewise, Bilgili et al. (2016a, b) infer that BEC mitigates the harmful effects of carbon dioxide on the environment in the USA. Another empirical study by Shahbaz et al. (2017) also reports the same findings for the USA.

Similarly, for BRICS economies, Dogan and Inglesi-Lotz (2017) confirm that BEC decreases the ratio of environmental pollution by reducing the emission of carbon dioxide. Besides, Baležentis et al. (2019) and Danish and Wang (2019) show that the level of CO₂ emission falls due to the environmental-friendly behavior of BEC for EU and BRICS economies, respectively. On the contrary, the second strand of scholars reveals the contradictive findings as compared with the first strand. For instance, Adewuyi and Awodumi (2017), deploying 3SLS technique, deduce that the adverse effects of CO₂ emission increase on the account of BEC in West Africa. Also, Solarin et al. (2018) gather that BEC affects environmental quality negatively by enhancing the emission of carbon dioxide in 80 developed and developing economies. Similarly, for G-7 economies, Shahbaz et al. (2018) conclude that the trend of environmental stress goes up due to an increase in CO₂ emission as more of biomass energy is consumed. Summarizing, again, the researchers present the scholarly disagreement regarding the BEC and CO₂ emission nexus as some of the scholars support the argument that BEC improves the environment quality by decreasing the emission of CO₂, while some other researchers disagree and report that environmental pollution increases on account of BEC. Further, there may not be a single study that explores the BEC and CO₂ emission nexus for Pakistan.

Methodology and data

Data and variable description

The present study has gathered data from WDI that covers the time period from 1975 to 2018. Regarding the dependent variable, CO₂ emission is used to proxy measured environmental quality, while two independent variables consist of clean energy. These comprise alternative and nuclear energy (% of total energy use) and combustible renewables and waste (metric tons of oil equivalent). Variables of electric power consumption (kilowatt-hour) and GDP are used as control variables. Electricity consumption includes all forms of energy whether clean and unclean. Moreover, these sets of energy variables are also being used together in the study of Maji (2015). All dependent and independent variables are retrieved from the WDI database. The data on dependent, independent, and control variables have been collected annually over the period 1975–2018 from WDI. We noted that data on some variables are available until 2014; however, we have extrapolated the data until 2018, to add few observations that will better serve the purpose in the case of our asymmetric model. Table 2 also gives the list of variable definition and measurement.

Table 2 Variable description

The mean of CO₂, ANE, CRW, EC, and GDP is 0.674 mt, 3.165%, 42.33 mt, 325.73 kWh, and 850.87, respectively, while the standard deviation is 0.189 mt, 0.491%, 8.838 mt, 129.89 kWh, and 165.02, respectively. The tendency of the variables, descriptive, and correlation statistics is offered in Table 3 and Fig. 1. Table 3 also shows the correlation matrix of the dependent and independent variables. The variables of ANE, EC, and GDP are positively correlated with CO₂, while the CRW is correlated with a negative manner on CO₂. In addition, EC and GDP are significantly negatively correlated to the RWC. Finally, there is also a significant positive relationship between EC and GDP.

Table 3 Summary statistics and correlation matrix

Fig. 1

Trend of the variables

Model and methodology

The main focus of this research is to estimate the impacts of different types of clean energy on CO₂ emission with special reference to Pakistan. To that end, we have selected two variables of clean energy: ANE and CRW. Hence, we have made the following simplest form of the model to analyze the connection between clean energy variables and CO₂ emission.

$$ {\mathrm{CO}}_{2,\mathrm{t}}={\beta}_0+{\beta}_1{\mathrm{ANE}}_{\mathrm{t}}+{\beta}_2.{\mathrm{CRW}}_{\mathrm{t}}+{\beta}_3{\mathrm{EC}}_{\mathrm{t}}+{\beta}_4{\mathrm{GDP}}_{\mathrm{t}}+{\mu}_{\mathrm{t}} $$

(1)

In Eq. (1), CO_2,t is our dependent variable, which represents annual carbon dioxide release in Pakistan. Our independent variables in the above model include alternative and nuclear energy (ANE_t), combustible renewables and waste (CRW_t), electricity consumption (EC_t), and national income of Pakistan (GDP_t), and μ_t is the error term. However, the above model, with any method, would only give us long-run estimates. To get short-run estimates, as well, we have followed the methodology suggested by Pesaran et al. (2001) which is commonly identified as ARDL. In this methodology, we have restated model (1) by including error correction specification into it as given below:

$$ {\Delta \mathrm{CO}}_{2,\mathrm{t}}={\delta}_0+{\sum}_{n=0}^p{\delta}_1{\Delta \mathrm{CO}}_{2,\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_2{\Delta \mathrm{ANE}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_3{\Delta \mathrm{CRW}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_4{\Delta \mathrm{EC}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_5{\Delta \mathrm{GDP}}_{\mathrm{t}-\mathrm{k}}+{\pi}_1{\mathrm{CO}}_{2,\mathrm{t}-1}+{\pi}_2{\mathrm{ANE}}_{\mathrm{t}-1}+{\pi}_3.{\mathrm{CRW}}_{\mathrm{t}-1}+{\pi}_4{\mathrm{EC}}_{\mathrm{t}-1}+{\pi}_5{\mathrm{GDP}}_{\mathrm{t}-1}+{\nu}_{\mathrm{t}} $$

(2)

The main benefit of this technique over other approaches is that it helps us to estimate short-run as well as long-run parameters in a single equation. Moreover, as this method can also consider the integrating properties of the variables, we do not need to worry whether our variables are I(0), I(1), or mixture of both. In the above Eq. (2), the short-run results are depicted by those coefficients which are connected with differenced (∆) indicators, whereas long-run results are represented by coefficients π₂₋π₅ normalized on π₁. To get accurate long-run outcomes, we must need to check co-integration among long-run variables and for that purpose, we rely upon the values of bounds F-test presented by Pesaran et al. (2001). To detect the joint significance of long-run estimates, we will check whether the calculated value of bounds’ test is large enough, i.e., larger than the values provided by Pesaran et al. (2001), or not; if calculated value is found to be higher than tabulated one, we can authorize the existence of co-integration. Though the values of bounds’ F-test presented by Pesaran et al. (2001) are suitable for data sets that have a large number of observations, we have trusted the values of Narayan (2005) due to our small sample size. However, if the critical values of F-statistics are not significant, then we will move to the next test of co-integration which is identified as error correction modeling. As far as this technique is concerned, error term from Eq. (1) is corrected by the aid of normalized long-run estimates. Then, we substitute this error correction term (ECM_t-1) with the lagged level variables in Eq. (2). The new equation is then estimated by employing the same number of lags. In order to check the significance of ECM_t-1, Pesaran et al. (2001) calculated new critical values of t test, just as they calculated critical values of F-test for joint significance of lagged level variables. If the value of ECM_t-1 is significant and negative, then we can confirm the presence of co-integration.

Next, in this study, our primary target is to check whether the influence of clean energy variables on the CO₂ emission is symmetric or asymmetric. For that purpose, we follow the Shin et al. (2014) methodology of non-linear ARDL and we break down energy variables into their positive (POS) and negative (NEG) components by following partial sum procedure:

$$ {{\mathrm{ANE}}^{+}}_{\mathrm{t}}={\sum}_{L=1}^t\Delta {{\mathrm{ANE}}^{+}}_{\mathrm{t}}={\sum}_{L=1}^t\max \left(\Delta {{\mathrm{ANE}}^{+}}_{\mathrm{t}},0\right) $$

(3a)

$$ {{\mathrm{ANE}}^{-}}_{\mathrm{t}}={\sum}_{L=1}^t\Delta {{\mathrm{ANE}}^{-}}_{\mathrm{t}}={\sum}_{L=1}^t\max \left(\Delta {{\mathrm{ANE}}^{-}}_{\mathrm{t}},0\right) $$

(4a)

$$ {{\mathrm{CRW}}^{+}}_{\mathrm{t}}={\sum}_{L=1}^t\Delta {{\mathrm{CRW}}^{+}}_{\mathrm{t}}={\sum}_{L=1}^t\max \left(\Delta {{\mathrm{CRW}}^{+}}_{\mathrm{t}},0\right) $$

(3b)

$$ {{\mathrm{CRW}}^{-}}_{\mathrm{t}}={\sum}_{L=1}^t\Delta {{\mathrm{CRW}}^{-}}_{\mathrm{t}}={\sum}_{L=1}^t\max \left(\Delta {{\mathrm{CRW}}^{-}}_{\mathrm{t}},0\right) $$

(4b)

$$ {{\mathrm{EC}}^{+}}_{\mathrm{t}}={\sum}_{L=1}^t\Delta {{\mathrm{EC}}^{+}}_{\mathrm{t}}={\sum}_{L=1}^t\max \left(\Delta {{\mathrm{EC}}^{+}}_{\mathrm{t}},0\right) $$

(3c)

$$ {{\mathrm{EC}}^{-}}_{\mathrm{t}}={\sum}_{L=1}^t\Delta {{\mathrm{EC}}^{-}}_{\mathrm{t}}={\sum}_{L=1}^t\max \left(\Delta {{\mathrm{EC}}^{-}}_{\mathrm{t}},0\right) $$

(4c)

Then, we entered these positive and negative shocks into linear ARDL model (2) as shown below:

$$ {\Delta \mathrm{CO}}_{2,\mathrm{t}}={\delta}_0+{\sum}_{n=0}^p{\delta}_1{\Delta \mathrm{CO}}_{2,\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_2{{\mathrm{ANE}}^{+}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\updelta}_3{{\mathrm{ANE}}^{-}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\updelta}_4{{\mathrm{CRW}}^{+}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_5{{\mathrm{CRW}}^{-}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_6{{\mathrm{EC}}^{+}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\updelta}_7{{\mathrm{EC}}^{-}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_6{{\mathrm{EC}}^{+}}_{\mathrm{t}-\mathrm{k}}+{\sum}_{n=0}^p{\delta}_6{\mathrm{GDP}}_{\mathrm{t}-\mathrm{k}}+{\pi}_1{\mathrm{CO}}_{2,\mathrm{t}-1}+{\pi}_2{{\mathrm{ANE}}^{+}}_{\mathrm{t}-1}+{\pi}_3{{\mathrm{ANE}}^{-}}_{\mathrm{t}-1}+{\pi}_4{{\mathrm{CRW}}^{+}}_{\mathrm{t}-1}+{\pi}_5{{\mathrm{CRW}}^{-}}_{\mathrm{t}-1}+{\pi}_6{{\mathrm{EC}}^{+}}_{\mathrm{t}-1}++{\pi}_7{{\mathrm{EC}}^{-}}_{\mathrm{t}-1}+{\pi}_8{\mathrm{GDP}}_{\mathrm{t}-1}+{\nu}_{\mathrm{t}} $$

(5)

The above equation is recognized as non-linear ARDL model as proposed by Shin et al. (2014). Once we estimated the non-linear model (5), then we will move towards the test of asymmetries. First, we will check the short-run joint impact asymmetries (∑δ_2k ≠  ∑ δ_3k, ∑δ_4k ≠  ∑ δ_5k, and ∑δ_6k ≠  ∑ δ_7k) by applying Wald test. If the cumulative sum of positive short-run shocks are not equal to the sum of negative shocks, we can indorse the occurrence of short-run asymmetry. Similarly, to check the long-run asymmetry ($\raisebox{1ex}{${\pi_2}^{+}$}\!\left/ \!\raisebox{-1ex}{${\pi}_1$}\right.$≠ $\raisebox{1ex}{${\pi_3}^{-}$}\!\left/ \!\raisebox{-1ex}{${\pi}_1$}\right.,\raisebox{1ex}{${\pi_4}^{+}$}\!\left/ \!\raisebox{-1ex}{${\pi}_1$}\right.$≠ $\raisebox{1ex}{${\pi_5}^{-}$}\!\left/ \!\raisebox{-1ex}{${\pi}_1\ $}\right.,\raisebox{1ex}{${\pi_6}^{+}$}\!\left/ \!\raisebox{-1ex}{${\pi}_1$}\right.$≠ $\raisebox{1ex}{${\pi_7}^{-}$}\!\left/ \!\raisebox{-1ex}{${\pi}_1$}\right.$) we will also apply Wald test. If the impacts of normalized long-run positive shocks are unlike from the impacts of negative shocks, for each energy variable, then it is a confirmation of long-run asymmetry.

Results and discussion

In an ARDL specification, it does not matter whether the variables are non-stationary processes I(1) or stationary processes I(0). Therefore, we confirm the order of the integration of variables through Phillips and Perron (PP) and Augmented Dickey-Fuller (ADF) test. The results described in Table 4 show that both ADF and PP tests recommend that all the variables are integrated of order one, except CO₂ emissions variable.

Table 4 Unit root test

In Table 5, using ARDL method, the long- and short-run coefficients in model 1 estimates are presented. The long- and short-run outcomes recommend that the coefficient of CRW is significant and inversely related to carbon emissions. While an adverse result is found in ANE and GDP, it has a significant positive influence on carbon emissions in the long run. Our primary focus, in clean energy and CO₂ emissions analysis, is on asymmetries. Therefore, in Table 5, we also examine the asymmetries in the nexus between clean energy and carbon emissions, both in short and long run. The outcomes specify that the estimate of the positive component of alternative and nuclear energy (ANE_t⁺) is significant at 10% level and equal to 0.117 but the coefficient of the negative alternative and nuclear energy shock (ANE_t⁻) is significant at 5% level and equal to 0.228. This indicates that in the long run the impact of alternative and nuclear energy on CO₂ is asymmetric, implying that the impact of positive shock of alternative and nuclear energy on CO₂ emission is different from that of negative alternative and nuclear energy shock. This implies that in short and long run, reductions in alternative and nuclear energy lead to an increase in non-renewable energy; non-renewable energy causes an increase in carbon emission in Pakistan and vice versa. Specifically, a 1% increase in alternative and nuclear energy leads to 0.117% increase in CO₂, but 1% decrease in alternative and nuclear energy leads to 0.228% increase in CO₂. It means that both positive and negative shocks are contributing to CO₂ emission in Pakistan. Nevertheless, the coefficient is small in magnitude; one possible reason is a relatively small share of ANE in the total energy mix in Pakistan that is 3.88%. According to Pakistan energy yearbook 2018, the energy share of contribution contains gas (29.02%), hydro (28.00%), thermal (21.91%), coal (12.44%), nuclear energy (3.88%), wind (3.59%), and solar (1.16%) (Hydrocarbon Development Institute of Pakistan 2018). The ANE is only a small share of total energy consumption and it increased very sluggishly, i.e., from 0.60% in 2010 to 3.88% in 2018. This finding is inconsistent with the work of Menyah and Wolde-Rufael (2010), Iwata et al. (2010), and Baek (2015), who suggested that nuclear energy helps to decrease CO₂ emission. This finding also indicates that the lack of facilities in nuclear energy consumption has largely increased the carbon emission, which enables the greenhouse gas problems and extensively disturbing the environments.

Table 5 Short-run and long-run estimates

Similarly, in short run, the impact of alternative and nuclear energy use on CO₂ is also asymmetric. The estimated coefficient of the ΔANE_t⁻ shock is negative and equal to 0.117; this implies that in short run, decreases in ANE lead to an increase in carbon emissions. A 1% decrease in ANE causes a 0.117% increase in carbon emissions. This result also implies that positive shock is also bigger in the short run compared with a negative shock. However, the ANE coefficient is also negative significant in time period (t − 1), suggesting 1 year before ANE had a negative significant influence on carbon emissions.

Furthermore, concerning the effect of combustible renewables and waste energy on carbon emissions, the findings reveal the asymmetric effect both in the short and long run. In the long run, the impact of both positive and negative renewables and waste energy shocks is negative, implying that a 1% improvement in renewables and waste energy leads to 2.088% decrease in carbon emissions and 1% decrease in renewables and waste energy consumption leads to 1.120% decrease in carbon emissions. This also informs that positive combustible renewables and waste energy shock are stronger than negative shock on carbon emissions. Similarly, in the short run, both negative and positive renewables and waste energy shocks have reduced the carbon emissions. This finding is consistent with the findings of Katircioglu (2015), Bilgili et al. (2016a, b), and Danish and Wang (2019), who reported that renewables and waste energy deplete carbon emissions, while the empirical finding is inconsistent with Adewuyi and Awodumi (2017), who noted that combustible renewables and waste have a positive influence on carbon emission in West Africa. The possible reason is renewables and waste energy perform like a clean energy source, which assists in controlling pollution by reducing CO₂ emissions in Pakistan. The overall result focused on clean energy bases on the fact that the government is currently employing the legal structure for clean energy in Pakistan.

Estimation results further reveal that in the long run, the impact of electric power consumption on CO₂ emissions is insignificant while asymmetric in short run. The coefficient of the ECT is negative and significant verifying the long-run equilibrium relationships among competing variables. Particularly, the ECT coefficient is 0.565 in ARDL and 0.759 in NARDL, respectively. This indicates that the speeds of convergence are 56.5% and 75.9%, suggesting the high speed of adjustment in Pakistan.

Diagnostic test results are reported in panel C in Table 5. The optimal lag order was selected (0,1,0,0,0) in the ARDL model. After estimating the linear ARDL and non-linear ARDL models, the diagnostic tests were carried out to check the validity of the estimated models. However, heteroskedasticity test, serial correlation test, normality test, and functional form tests were used and the results revealed that they fail to reject the hypotheses of no heteroskedasticity, no serial correlation, normal residuals, and correct functional form in the ARDL and NARDL model. The diagnostic tests infer that ARDL and NARDL estimated are usable and do not have any inferential problem. In addition, adjusted R² is equal to 0.985 in ARDL and 0.986 in NARDL, indicating a good fit. Also, in asymmetric co-integration tested by using the F-test, the result confirmed that asymmetric co-integration exists in the short and long run. We have applied non-linear model because, in real-world scenario, symmetry in the effects of variables is hardly found. Moreover, literature does not suggest any of such tests that would be used to detect non-linearity in the data before applying NARDL. Few instances include studies like Apergis (2014), Cosmas et al. (2019), Karasoy (2019), and Awodumi and Adewuyi (2020); all these studies have checked the asymmetric impacts of different variables on environmental quality without the pre-estimation testing. The literature also suggests the post-estimation non-linearity test. Therefore, we have performed the Wald tests, after estimating our model, which confirms the existence of asymmetry in the effects of clean energy variables. In the case of nuclear energy and electricity consumption, Wald test confirms the presence of short-run impact asymmetries, in the effects of positive and negative parts of respective variables. As far as long-run asymmetry between positive and negative shocks is concerned, Wald test rejects it. On the other side, Wald test approves the non-linearity in the long-run positive and negative shocks of renewables and waste energy variable, whereas Wald test is insignificant in the case of short run.

Furthermore, we also performed CUSUM and CUSUMSQ tests to verify the stability of the parameters in the ARDL and NARDL model. However, the plot of CUSUM test in the ARDL model is unstable nevertheless. As far as this study is concerned, our primary objective is to check the asymmetry in the effects of clean energy variables. Therefore, this will not affect the validity of our non-linear results. Interestingly, the tests infer that computed NARDL model parameters are stable over the time period and the estimated model is reliable in Fig. 2.

Fig. 2

The results of CUSUM test for the ARDL and NARDL model

Conclusion and policy implications

This study observes the asymmetries in the link between clean energy consumption and carbon dioxide emissions in Pakistan using the non-linear or asymmetric ARDL model approach for the period of 1975–2018. The results from the estimated asymmetric ARDL model show that asymmetries exist in the relationships among ANE and carbon emissions in Pakistan, both in the short run and long run; the effect of negative shocks is dissimilar from that of positive shocks. Similarly, CRW affects carbon dioxide emissions, indicating that carbon dioxide emissions are affected by positive and negative energy shocks in the long run, while negative and positive energy shocks affect carbon dioxide emissions in the short run differently. Although short- and long-run results suggest an insignificant positive and negative relationship between electric power consumption and carbon emissions, the non-linear ARDL model results confirm that the asymmetry supposition can be supported in Pakistan. Indeed, asymmetric variations in clean energy are essential in the case of Pakistan, with estimated economic signs.

These empirical outcomes suggested that asymmetric co-integration exists in the model. This study is consistent with the study of Karasoy (2019), who noted that clean energy has an asymmetric effect on carbon emissions in Turkey. Overall, positive and negative shocks in ANE increase carbon emissions, whereas positive and negative changes in CRW consumption decrease carbon emissions in Pakistan. These results are valid and noticeable in Pakistan.

A rise in unclean energy consumption worsens the environment of Pakistan in long run, while an innovation of clean energy improves the environmental quality of Pakistan. The government should encourage clean energy for a healthier environment. Government and policymakers should give the long-run incentives and pursue the policies which would help to increase clean energy production in Pakistan. We urge that the governments of Pakistan should import new clean energy production technology. The government should take positive steps to change non-renewable energy to clean energy consumption that will not only reduce its energy dependency but also be a positive step towards combating global warming. Policymakers should pay more attention to public awareness about clean energy and its role in clean and green Pakistan. Regarding the limitation of the study, future research should examine the asymmetric effects of fossil fuel energy consumption on environmental pollution in Pakistan. Therefore, further empirical research in this area will appear fruitful in the policy context.